Anthropic's alignment science lead, Evan Hubinger, has stated that the company behind the Claude AI assistant genuinely believes artificial intelligence could pose an existential threat to humanity, potentially within the next ten years. In a public response to the recent resignation of fellow researcher Jacob Coxon, Hubinger wrote that he personally estimates a greater than 10% chance of AI causing human extinction in the coming decade. He acknowledged that while Anthropic is making its best effort, the company does not yet have a definitive plan to solve the alignment problem for superintelligent systems.
Hubinger pointed to the risk of recursive self-improvement, a scenario in which AI systems rewrite their own code to become progressively more capable, as a primary concern. He cited Anthropic's own risk assessment, which currently labels the immediate danger from existing models as low, but noted that the trajectory toward superintelligence remains a serious worry. Coxon, who stepped down from his role at Anthropic citing safety concerns, criticized both Anthropic and OpenAI for what he described as a lack of responsible action. He warned that these systems will soon become superhuman in their capabilities, able to hack any system, transform any field overnight, and acquire real power and resources.
Coxon explained that Anthropic is pushing forward despite these risks because it is locked in a competitive race to reach artificial general intelligence first. He suggested that while OpenAI may not be taking the issue seriously enough, Anthropic understands the civilizational stakes and feels compelled to act responsibly itself, even if that means accepting significant risk. The exchange has generated considerable debate across social media and political circles.
Senator Bernie Sanders of Vermont weighed in, arguing that AI threatens the economy, privacy, democracy, and the well-being of children. He called on Congress to stand up to what he termed big tech oligarchs and protect the American people. Conversely, technology reporter Taylor Lorenz dismissed Hubinger's comments as sanctimonious doomer statements that could lead to poorly conceived legislation. She challenged him to provide concrete proof of the dangers rather than vague warnings that might result in harmful policy.
The discussion comes at a pivotal moment for Anthropic, which is expected to go public this year with a valuation estimated at over $2 trillion. Some observers have speculated that the company's stark warnings about AI risk could be a strategic move designed to shape future regulation in its favor, following a business model of creating a problem and then selling the solution. If AI regulation is introduced, Anthropic could be positioned to help draft the rules in ways that benefit its own operations.
Anthropic's CEO, Dario Amodei, has previously stated that AI could eliminate half of all entry-level white-collar jobs in the coming years, comments that were widely interpreted as part of fundraising efforts. The broader AI industry continues to fuel significant growth in the US economy, though many analysts are warning of a potential bubble that could lead to substantial market disruption. The ongoing debate highlights the tension between rapid technological advancement and the need for adequate safety measures, a balance that remains unresolved as companies race toward increasingly powerful systems.